Qinxun Bai http://cs-people.bu.edu/qinxun/ Email: qinxun@bu.edu Objective Research oriented industrial positions in machine learning and computer vision. Education Boston University Ph.D. Candidate in Computer Science Thesis: “Differential Geometric Approach to Classification” Advisors: Prof. Stan Sclaroff and Prof. Steven Rosenberg The National Laboratory of Pattern Recognition, Chinese Academy of Sciences M.S. in Pattern Recognition and Intelligent Systems Thesis: “Robust 3D Scene Reconstruction” Advisor: Prof. Yihong Wu Tsinghua University B.Eng. in Electronic Engineering Professional Experience Image and Video Computing Group, Boston Univ. Research Assistant Differential Geometric Framework for Regularization Advisors: Prof. Steven Rosenberg and Prof. Stan Sclaroff New geometric perspective on overfitting, First regularization approach that exploits the geometry of a robust class probability estimator for classification, Unified framework for both binary and multiclass cases, Compares favorably with existing regularization methods, Ongoing applications include Deep Neural Networks, multi-label classification, and unsupervised learning. http://cs-people.bu.edu/qinxun/geo/geo.html Bayesian Asymptotics for Stochastic Optimization Advisors: Prof. Henry Lam and Prof. Stan Sclaroff Bring Bayesian statistics to stochastic global optimization, Global convergence guarantee without convexity condition, Favorable convergence rate compared with standard SGD, A new and promising method for online ensemble learning. http://cs-people.bu.edu/qinxun/boe/boe.html Online Ensemble Learning for Visual Tracking Advisors: Prof. Stan Sclaroff and Prof. Margrit Betke A new Bayesian method for ensemble tracking, Strong occlusion handling when coupled with a fine grid appearance model, Outperforms previous ensemble tracking methods. http://cs-people.bu.edu/qinxun/ret/ret.html Boston, MA 09/2010 - present Beijing, China 09/2008 - 06/2010 Beijing, China 09/2002 - 06/2006 Boston, MA 09/2010 - present Microsoft Research Cambridge Research Intern Mentors: Sebastian Nowozin and Daniel Tarlow A PatchMatch-based idea to speed up the Markov chain Monte Carlo (MCMC) for large random field models. The National Laboratory of Pattern Recognition, Chinese Academy of Sciences Research Assitant Advisor: Prof. Yihong Wu New method for extracting structured information (collinearity and coplanarity) for scene reconstruction, Used in the NSF China key project for complex and large scale scene reconstruction. The National Laboratory of Pattern Recognition, Chinese Academy of Sciences Software Engineer Mentor: Prof. Stan Z. Li Built a 3D face synthesis system that generates 3D faces from a single frontal face image, The system served for the Face Authentication System for Beijing 2008 Olympics, generating multi-view training data. Cambridge, UK Summer 2014 Beijing, China 09/2008 - 06/2010 Beijing, China 09/2006 - 06/2008 Skills Proficiency in C++ and Matlab with 10 years experience. Familiar with Lua. Layer module design experience for deep learning based on Torch. Publications Qinxun Bai, Steven Rosenberg, Zheng Wu and Stan Sclaroff. “Class Probability Estimation via Differential Geometric Regularization.” To appear in Proc. of International Conference on Machine Learning (ICML), 2016. Qinxun Bai, Henry Lam and Stan Sclaorff. “A Bayesian Approach for Online Classifier Ensemble.” Under review, arXiv preprint arXiv:1507.02011, 2015. Qinxun Bai, Henry Lam and Stan Sclaorff. “A Bayesian Framework for Online Classifier Ensemble.” In Proc. of International Conference on Machine Learning (ICML), 2014. Qinxun Bai, Zheng Wu, Stan Sclaroff, Margrit Betke and Camille Monnier. “Randomized Ensemble Tracking.” In Proc. of the IEEE International Conference on Computer Vision (ICCV), 2013. Qinxun Bai, Yihong Wu, and Lixin Fan. “PCA-based Structure Refinement for Reconstruction of Urban Scene.” In Proc. of the IEEE International Conference on Image Processing (ICIP), 2010. Zhen Lei, Qinxun Bai, Ran He and Stan Z. Li. “Face Shape Recovery from a Single Image Using CCA Mapping between Tensor Spaces.” In Proc. of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2008. Professional Service Reviewer for: Computer Vision and Image Understanding Neurocomputing ICML ’16 IJCAI ’16 AAAI ’14 IEEE Transactions on Image Processing (external) ICCV ’15 (external) CVPR ’12, ’13, ’16 (external) Teaching fellow at Boston University: CS480/680 Introduction to Computer Graphics, Spring 2011 CS542 Machine Learning, Spring 2013 CS235 Algebraic Algorithms, Fall 2013, Spring 2014 CS132 Geometric Algorithms, Spring 2015, Fall 2015 References Stan Sclaroff, Boston University Associate Dean of the Faculty, Mathematical & Computational Sciences Professor of Computer Science 111 Cummington Mall, Rm 279 Boston, MA 02215 sclaroff@bu.edu Steven Rosenberg, Boston University Professor of Mathematics & Statistics 111 Cummington Mall, Rm 248 Boston, MA 02215 sr@bu.edu Henry Lam, University of Michigan Assistant Professor of Industrial and Operations Engineering 1205 Beal Ave. Ann Arbor, MI 48109 khlam@umich.edu